HW week 8

HW week 8

by Amy -
  1. What are the different ways to account for SES in an analytic model when investigating racial/ethnic health disparities? (Hint: you should have three options). Discuss the interpretations/implications of each approach as it relates to the interest in understand health disparities by race/ethnicity. Draw a DAG for each option and reference it in your response (you do not have to post this!).
    1. Using generalized estimating equations with log link functions to estimate risk ratios for disease by race/ethnicity (and other factors), in which measures of SES are considered confounders and controlled for in the model.
    2. Developing a series of logistical regression models in which SES is included as a mediator, as in Lorch, acting a factor on the causal pathway between race/ethnicity (exposure) and disease so that the crude association between race/ethnicity and disease can be compared to the association when SES is included in the model.
    3. Using multilevel regression analysis (MLRA), as in Merlo, to show that the association between race/ethnicity and disease is a product of both individual and contextual level (e.g. neighborhood or country) differences.
  2. Think about multilevel influences on a health outcome of interest to you. Discuss how you would study this, including measurement and analytic approaches you would use to account for exposures across multiple levels.

This is a bit of a stretch, but I’m trying to apply this to my research interest in the effects of quality of program management in global health programs on health outcomes. This interest is based on the assumption that health outcomes are not as good as they can be in developing countries. It might be interesting to look at differences in a health outcome (e.g. maternal or infant mortality) by country (as the exposure) using factors of health delivery as mediators. These mediators could include: local budget per person, donated funds per person, number of HCPs per person, quality of facilities, and quality of program management (among others). Similar to the analysis done in the Lorch paper, this would allow me to understand what part of the disparity in health outcome is due to these different aspects of the system.

  1. Respond to one other person's post on the forum with a comment or suggestion.

TBD

Re: HW week 8

by Emily -

Could you also apply the multilevel approach by measuring individual, organizational, and societal factors as they impact an outcome? This could look something like an individuals contribution to a program at a health center, the MOH's support (or lack of support) of the program (new implementation, changes in implementation), and commonly held knowledge among the users of the program. I am having a hard time thinking of a natural experiment, but when linked to the role out of a new program in a geographically defined area it could be possible and interesting.

Re: HW week 8

by Maricianah -

HI Amy

 

I agree with you that a multi-level regression analysis as described by Merlo et al,may be a great way to approach your question. Given the scope of your study and for the outcomes you are looking at (infant mortality/morbidity)- you are likely going to do an ecological study and these countries could cluster by east and southern region (ESARO), west and central Africa, and then north Africa. Alternatively, historically, you could cluster them by who colonised them – I have found that there are significant disparities between Anglophone and francophone Africa and even worse when you compare with portugese Africa. Typically the WHO guidelines are made in English,  few copies in French and even much fewer in Portuguese and for that simple reason- implementation and standardisation of important life saving interventions is a real challenge.

 

As a framework for the other levels- I agree with you around the costing- one of the things we did while I was in WHO – was just look at what was the intercountry budgetary allocation for child health commodities or family planning commodities, differences in number of documented policies enacted around child health and then create a score. In hind sight- and I could be wrong – these different levels of health systems could have been coded and placed into a multi-level regression analysis

 

Certainly very exciting to even think about this 

 

Maricianah

Re: HW week 8

by Christine Dehlendorf -

I agree with everyone's comments that your multilevel contextual factors are quite interesting and that thinking about what factors might mediate findings of associations by geographies or system is an important next step after identifying contextual assocaitions. Other things to consider include things like community-wide technological literacy and immigration (both in and out).

For modeling SES, considering SES a confounder, or "third variable" is definitely an option, with GEE or regular regression models. Also think about the possibility for SES to be an effect modifier.